Biology Reference
In-Depth Information
strikingly similar expression profiles. Most importantly, the
study was able to assign eight unannotated genes to the
regulation of pathways such as sterol metabolism, protein
synthesis and mitochondrial function. Furthermore, the
observation that gene expression profile in response to drug
treatment phenocopies the loss-of-function profile of its
target facilitated the identification of Erg2p, a sterol isom-
erase, as a novel target of the drug dyclonine
[231]
.
Another study in Saccharomyces cerevisiae
[240]
analyzed the functional relationship between kinases and
phosphatases by generating genome-wide expression
signatures for 150 deletion mutants. Gene expression
signatures were also generated for pairs of genes (kinases
and phosphatases) that exhibit synthetic genetic interac-
tions with the aim of investigating the mechanisms under-
lying the redundant relationships. The results of this study
concluded that there are three types of redundant connec-
tion: (1) complete redundancy, where the two genes in
a synthetic genetic interaction regulate the same set(s) of
genes to an equal extent, such that the single mutants show
no significant changes, but an effect on expression of
regulated gene set(s) is seen only in the double mutant; (2)
quantitative redundancy, where the two genes in a synthetic
genetic interaction regulate the same set(s) of genes but to
a quantitatively different extent. Here one of the single
mutants shows no significant effect but the other does, and
the effect on the expression of the regulated gene set(s) is
amplified in the double mutant; and (3) mixed epistasis,
where the two genes in a synthetic genetic interaction
regulate some of the same gene set(s) via either complete or
quantitative redundancy, while other gene set(s) behave in
a completely different way. Mixed epistasis reflects only
a partial overlap in function of the two genes in the
synthetic genetic interaction. The authors concluded that
such gene pairs share additional regulatory associations,
such as inhibition of one by the other, and that mixed
epistatic relationships provide the mechanisms to achieve
signaling specificity in a context-dependent manner.
Importantly, mixed epistasis was found to be the most
common redundant relationship in signaling networks.
The Connectivity Map (Cmap)
[241,242]
, identified
functional connections between drugs, genes and diseases
from expression profiles in a compilation of genome-wide
expression data from cultured human cells treated with
either bioactive small molecules or genetic perturbations.
Cmap incorporates pattern-matching algorithms that
decode differential gene expression data into functional
relationships between drugs, genes and diseases to generate
testable hypotheses. Again, the underlying assumption of
Cmap is that common gene expression changes reflect
functional connectivity between the gene products targeted
by either small molecule or various genetic perturbations.
Functional connectivity is expected to reflect the role of
gene products in a common biological process, in particular
components of a specific signaling pathway. Cmap was
used to identify the target of two previously uncharac-
terized natural products (celastrol and gedunin) that had
inhibitory activity towards androgen receptor activity, with
implications for the treatment of prostate cancer
[243]
.
Gene expression signatures for each of the drugs were
generated and used to search the Cmap database for similar
gene expression patterns. The gene expression signatures of
celastrol and gedunin were most similar to the signatures of
inhibitors of the chaperone HSP90. This finding predicted
that HSP90 was the most likely target of celastrol and
gedunin activity, a hypothesis that was tested and validated
experimentally. In another study, rapamycin was found to
reverse the effects of resistance to the glucocorticoid
dexamethasone in acute lymphoblastic leukemia (ALL)
[244]
.
A number of publicly available compendia of gene
expression profiles are available for data mining purposes,
including the Global Cancer Map
[245]
, Gene Expression
Atlas
[246,247]
, and Oncomine Cancer Profiling Database
[248]
.
The overall logic of establishing connectivity based on
gene expression signatures of RNAi treated cells is simple
and schematically presented in
Figure 5.2
. The current
challenge is to go beyond proof-of-principle studies and
establish robust experimental protocols and computational
tools that will allow large-scale implementation. This
presents three main challenges: (1) the generation of gene
expression signatures for every biological state of interest
(disease, cellular activity such as cell division, response to
drugs or genetic perturbation); (2) a cost-effective high-
throughput platform for screening genetic perturbations or
small molecule treatment using gene expression signatures;
and (3) the development of computational tools for data
analysis.
Gene expression signatures serve as molecular surro-
gates for biological states, are composed of tens to
hundreds of genes, and are distinct for different biological
states
[235,249]
. The ability to identify a gene expression
signature that can serve as a quantitative molecular
phenotype for a specific biological state holds great
promise for the development of high-throughput small-
molecule or RNAi screens using the signature of interest as
the readout. The first gene expression signature-based
screening (now called gene-expression-based high-
throughput screening, GE-HTS) was conducted
[249]
to
identify small molecules that induce the differentiation of
acute myeloid leukemia cells. Using gene expression
analysis of primary cells from patients and unaffected
individuals identified a number of differentiation-
correlated genes to generate a five-gene signature for
leukemia cell differentiation. This signature was then used
to screen a library of 1739 bioactive small molecules
to identify those that
induce the expression of
the
